Skip to main content
NCBI home page
Cognitive Neurodynamics logoLink to Cognitive Neurodynamics
. 2022 Jan 24;16(5):1107–1121. doi: 10.1007/s11571-021-09765-z

Hybrid brain model accurately predict human procrastination behavior

Zhiyi Chen 1,2,#, Rong Zhang 1,2,#, Jiawei Xie 3, Peiwei Liu 4, Chenyan Zhang 5, Jia Zhao 1,2, Justin Paul Laplante 6, Tingyong Feng 1,2,✉
PMCID: PMC9508313  PMID: 36237406

Abstract

Procrastination behavior is quite ubiquitous, and should warrant cautions to us owing to its significant influences in poor mental health, low subjective well-beings and bad academic performance. However, how to identify this behavioral problem have not yet to be fully elucidated. 1132 participants were recruited as distribution of benchmark. 81 high trait procrastinators (HP) and matched low trait procrastinators (LP) were screened. To address this issue, we have built upon the hybrid brain model by using hierarchical machine learning techniques to classify HP and LP with multi-modalities neuroimaging data (i.e., grey matter volume, fractional anisotropy, static/dynamic amplitude of low frequency fluctuation and static/dynamic degree centrality). Further, we capitalized on the multiple Canonical Correlation Analysis (mCCA) and joint Independent Component Analysis algorithm (mCCA + jICA) to clarify its fusion neural components as well. The hybrid brain model showed high accuracy to discriminate HP and LP (accuracy rate = 87.04%, sensitivity rate = 86.42%, specificity rate = 85.19%). Moreover, results of mCCA + jICA model revealed several joint-discriminative neural independent components (ICs) of this classification, showing wider co-variants of frontoparietal cortex and hippocampus networks. In addition, this study demonstrated three modal-specific discriminative ICs for classification, highlighting the temporal variants of brain local and global natures in ventromedial prefrontal cortex (vmPFC) and PHC in HP. To sum-up, this research developed a hybrid brain model to identify trait procrastination with high accuracy, and further revealed the neural hallmarks of this trait by integrating neuroimaging fusion data.

Supplementary Information

The online version contains supplementary material available at 10.1007/s11571-021-09765-z.

Keywords: Procrastinators, Fusion data, Machine learning, Multiple canonical correlation analysis, Diagnostic biomarkers

Introduction

Procrastination refers to a near ubiquitous scenario: peoples are frequently prone to postpone intended tasks until the deadline has arrived (Steel 2007). Despite no significant physiological pains, a certain number of individuals are suffering from problematical procrastination (generally refers to trait procrastination behavior). This undoubtedly engenders an array of negative life outcomes including poor cognitive abilities, low subjective well-beings, and unhealthy physical conditions (e.g., obesity) (Sirois 2007). Furthermore, the high prevalence of trait procrastination found in previous investigations is worrisome as well: 80–95% of students have reported frequent academic procrastination, and more than 15% of adults were found as problematic procrastinators (Steel 2007). Notably, trait procrastination is not only defined as a behavioral problem but also receives increasing attentions to link with a wide variety of neuropsychiatric disorders, such as major depression and anxiety (Stöber and Joormann 2001; Walsh and Ugumba-Agwunobi 2002).

Thus far, much research has provided initial evidence that the brain structural underpinning of trait procrastination is characterized by the prominent anomalies in self-control network (e.g., dorsolateral prefrontal cortex (dlPFC) and anterior cingulate cortex (ACC)), emotional regulation network (e.g., insula and orbital frontal lobe (OFC)), and episodic prospection network (hippocuampus (PHC) and ventromedial prefrontal cortex (vmPFC)) (Chen et al. 2019; Hu et al. 2018; Liu and Feng 2017). In addition, neurofunctional substrates of trait procrastination have been also probed in previous works, and it demonstrated the dysfunctional intra-connection of the default mode network and aberrant spontaneous brain activation of PHC/vmPFC during resting-state in individuals with trait procrastination (Wu et al. 2016; Zhang et al. 2016). Furthermore, a task-related fMRI study aiming to unveil the neural mechanism of trait procrastination demonstrated that the problematic procrastinators may be attributed to be vulnerable PHC-striatal circuit (Zhang et al. 2019a; b). Taken together, though existing finings enriched our understanding of trait procrastination by using either brain structural features or neurofunctional hallmarks, identifying trait procrastination basing on multivariable brain features has not been as directly studied.

Unlike the canonical univariate analysis conducted for brain features, a growing number of studies reap huge fruits from adopting multivariable analysis for fusion data to diagnose mental illness or neuropsychiatric disorders, especially when building a hybrid brain model (Antonucci et al. 2019). Typically, the fusion data contains brain information from at least two neural modalities, such as structural and functional schemes. As for the brain structural features, the volumes (GMV) of grey matter tissue and fractional anisotropy (FA) of white matter tissue are the mainstreams untill now (Buckner et al. 2009; Cercignani et al. 2001; Li et al. 2021). Meanwhile, local spontaneous amplitude of low frequency fluctuation (ALFF) of voxels and global connectome of voxels (i.e., Degree Centrality, DC) are strikingly common configurations of brain functional features, which depicts the brain information in local and connectome profiles respectively (van den Heuvel and Sporns 2013; Yang et al. 2007; Zou et al. 2008). Furthermore, given that the spontateous brain activity fluctuates are highly dynamic, the insights into the neurofunctional features have been thus extended from static state to the dynamic state in the recent studies (Liao et al. 2015; Casorso et al. 2019). Encouragingly, identifying and even diagnosing neuropsychiatric disorders by using fusion features of aforementioned brain modalities has been found to outperform sole modality information, such as diagnosing schizophrenia (SZ), obsessive–compulsive disorder (OCD), and autism spectrum disorder (ASD) (Sui et al. 2013a, b; Itahashi et al. 2015; Kim et al. 2015).

In addition, the data-driven multiple canonical correlation analysis and  joint independent component analysis (mCCA + jICA) are broad-certified suitable to provide interpretable results for brain fusion data. This conjunction algorithm could largely detected the shared information derived from the multi-modalities for each independent component, thus enabling the comprehensive identification of the potential brain abnormities available and productive (Sui et al. 2012). Technically speaking, it evaluated the decomposed components based on the mixing profiles which overcomes the pitfalls found in the either one (Sui et al. 2012, 2013a; b). High-performance of the mCCA + jICA algorithm have been attested in the widespread utilization by using multi-modalitied biomarkers, particularly in the diagnosis of neuropsychiatric diseases, such as bipolar disorder and schizophrenia (Sui et al. 2011, 2013a; b). In this vein, by fusing both brain structural features and functional architectures, the mCCA + jICA could provide interpretable results for the modal-joint and modal-specific components.

In the present study, we built upon a hybrid brain model by using these fusion brain data, including structural modalities (GMV × FA), static functional properties (static fALFF of local activation × static DC of global connectome), and dynamic functional schemes (dynamic fALFF of local activation × dynamic DC of global connectome). Afterwards, the multi-voxel pattern classification (MVPA) of machine learning was used to train and test this hybrid brain model for discriminating between HP and LP. Specifically, both the binary support vector machine (SVM) of the single kernel function and the L1-Multiple Kernel Learning (L1-MKL) classifier were adopted for training classifers so as to compare performance of fusion brain data to each brain modality. Finally, to obtain neuropathological pattern of these fusion features in identifying HP and LP, the N-way mCCA + jICA model was built as well.

The major hypothesis for this work has been pre-registered beforehand at Open Science Framework (OSF, https://doi.org/10.17605/OSF.IO/G5S9Q). Major goals of the current study are two-fold: one is to identify trait procrastination accurately by using this hybrid brain model; another one is to adopt mCCA + jICA model revealing the neural markers of trait procrastination based on the fusion brain data. Thus, we hypothesized that the hybrid brain model could outperform any models using sole brain modality in identifying this trait. Further, we posited that the fusion neural components for emotional regulation, self-control, and episodic prospection networks might be found as biomarkers of trait procrastination.

Methods and materials

Participants

According to the quantifiable examination and benchmark distribution (Tuckman 1991; Ferrari 1992; Rebetez et al. 2014) (see Supplementary Information, SI), 81 high trait procrastination (HP) participants (procrastinators) were screened (32 Males; Age, M: 21.19, S.D: 1.85, Range: 18–25). Meanwhile, 81 participants were enrolled as healthy control with low trait procrastination (LP), matching for the age, gender and other demographic information (32 Males; Age, M: 20.88, S.D: 2.16, Range: 18–25). All of the participants were confirmed for free of the neuropsychiatric diseases using the both Axis I and Axis II of DSM-IV in the posteriori interview (First et al. 1996, 2001). Details for demographic information and aprior identification for statistical power can be found in Table 1 and SI Methods.

Table 1.

Overview of the demographic information for the participants of this study

HP P-val LP P-val P-val
Male Female Male Female
Num 32 49 – 32 49 – –
Age 20.3 (1.7) 20.0 (1.9) .43 20.0(1.7) 19.8 (1.4) .35 .29
Educ 14.1 (1.8) 13.9 (2.0) .43 14.0 (1.7) 14.2 (1.0) .70 .29
BMI 20.9 (3.2) 21.4 (3.8) .51 20.9(4.4) 21.4(3.8) .21 .88
BDI 20.5 (6.2) 19.0 (4.8) .22 17.3 (3.8) 18.1 (4.2) .51 .01*
TA 59.0 (9.4) 55.1 (7.4) .57 53.7 (8.9) 52.5 (8.0) .38 .01**
NEO-C 34.7 (5.0) 36.7 (5.8) .13 41.0 (4.4) 42.8 (5.7) .48 .00***
NEO-E 38.3 (8.7) 39.2 (7.8) .64 38.3 (8.7) 38.1 (6.9) .68 .59
NEO-N 39.9 (8.6) 36.0 (7.5) .05 36.0 (7.5) 35.3 (8.4) .24 .12
NEO-A 39.4 (5.7) 39.8 (5.7) .74 39.9 (4.0) 41.2 (5.4) .20 .21
NEO-O 43.7 (6.3) 41.1 (6.2) .07 42.7 (6.6) 41.0 (5.4) .38 .70
PPS 56.6 (1.2) 55.2 (1.2) .51 36.0 (0.0) 35.8 (0.4) .58 .00***
Open in a new tab

“Num.” indicates the number of the corresponding participants. The cell of the 2th—3th and 5th—6th show the mean value and the corresponding standard deviation within the brackets. p-Val indicates the p-value derived from the the rank-sum Mann–Whitney U test, whilst the 4th and 7th columns present the p-value for the test on gender’s differences of HP and LP respectively. The last column reports the p-value of the examination for the differences between the HP and LP. BMI Body Mass Index; BDI indicates the diagnosis of the depression using the Becker Depression Inventory; TA Trait Anxiety; NEO-C Conscientiousness of big-five personality; NEO-E Extraversion of big-five personality; NEO-N Neuroticism of big-five personality; NEO-A Agreeableness; NEO-O Openness of big-five personality; PPS Pure Procrastination Scale. Asterisk indicates that the p-value have attain the significant level, *p < 0.05; **p < 0.01; ***p < 0.001

Identification of the HP

The Pure Procrastination Scale (PPS; Steel 2010) was used as quantifiable measurement for identification of HP and LP. PPS served as a powerful and productive tool to rapidly characterize one’s symptom towards procrastination (Svartdal and Steel 2017). This tool was designed by Steel (2010) to negate the pitfalls of conventional scales of the procrastination using “blurry” cognitive components, and has been validated robust in cross-cultural contexts (Steel 2010; Rozental et al. 2014; Svartdal and Steel 2017; Rebetez et al. 2018). Participants whose summed scores were greater than 49 points on this scale were identified as HP. More details for identifying HP/LP and data cleaning can be found in SI Methods. PPS showed satisfactory reliability in the current study (Cronbach's α = 0.87; Actual Variance Error [AVE] = 0.61; Composite Reliability [C.R] = 0.83).

Protocol of the scan

The current study acquired both structural and functional MRI images in the same scanner: 3 Tesla SIEMENS MRI Trio series (Siemens Medical Department, Erlangen, Germany). Prior to the scanning, all the participants were informed to stay awake with no specific thoughts in the scanner. To constrict heal-motion movements, the sanitary foam padding was used.

sMRI parameter

The high-resolution T1*-weighted images for each participant were acquired with 32-channel orthogonal head coil from this scanner using the magnetization-prepared rapid gradient echo (MPRAGE) pulse sequence, with 128 slices of the contiguous sagittal maps at the thickness of 1.33 mm (Time Repeat [TR] = 2530 ms; Time Echo [TE] = 3.39 ms; Time Interval [TI] = 800 ms; flip angle = 7°; Field of Visual [FoV] = 256 × 256 mm2; BW[Hz/Px] = 210; voxel size = 1.00 × 1.00 × 1.33 mm3).

For the collection of the diffusion weighted images (DWI), all the participants were scanned leveraging the cardiac-gated Spin Echo Diffusion Weighted Echo Planar Imaging (SE-DW-EPI) sequence with duration of 8 min 11 s. All of the scanning parameters were in line with previous canonical researches (Alexander et al., 2007; Basser et al., 1994). More details were denoted as below: TR = 7200 ms; TE = 104 ms; Native Resolution = 2.5 × 2.5 × 3.0 mm3; Slices = 49 axial maps; Matrix = 128 × 128 (RO × PE); FoV = 230 × 230 mm2; factor b maximum = 1000 s/mm2; b0 = 0 s/mm2; non-colinear diffusion directions = 64.

fMRI parameter

As to the acquisition of the brain functional images, the T2*-weighted maps were scanned at intervals, with odd slices for the superior scanning (EPI sequence, slices = 32; thickness of slice = 3 mm; inter-slice gap = 0.6 mm; TR = 2000 ms; TE = 30 ms; flip angle = 90°; matrix size = 64 × 64 mm2). To obtain more high-resolution functional images, this scanning took 10 min for a total of 300 volumes for each participant. During the scanning, all of participants were informed to keep their eyes open for a rest, but not to sleep or thinking towards special businesses.

Head-motion scrubbing

Given the increasing concerns for the artifacts stemming from the arbitrary head motion, the ICA-AROMA algorithm was carried out for more rigorous correction towards head motion (Pruim et al. 2015a, b). In detail, this strategy referred to a ICA-based scrubbing for head-motion artifacts for resting-state functional MRI data. Drawing upon probabilistic ICA of Multivariate Exploratory Linear Decomposition into Independent Components (MELODIC), a set of independent spatial and corresponding temporal ICs were yielded. Subsequently, these ICs re-constructed from blind data sources of fMRI were further determined to link ICs for head-motion component. Finally, these identified ICs which linked to head motion were adjusted by linear regression of ordinary least squares (OLS) method. ICA-AROMA correction was implemented with DPABISurf V1.1 (http://rfmri.org/dpabi) in part (Yan et al. 2016).

Preprocessing and estimations of brain features

Preprocessing of neuroimaging data was highly in accordance with common protocols (see more details in SI Methods) (Esteban et al. 2019). Notably, instead of conventional smoothness of Gaussian kernel, the 3D denoising of wavelet transformation was utilized not only for ameliorating of the noise-signal ratio (NSR) but also for maintaining high-resolution of spatial information meanwhile (Friston et al. 2000; Wang et al. 2005; Khullar et al. 2011). Details for estimations of brain structural features (i.e., GMV and FA) and functional features (i.e., static/dynamic fALFF and DC) are illustrated in Fig. 1 and SI Method.

Fig. 1.

Fig. 1

Open in a new tab

Diagram of the calculation for both the dynamic fALFF (fractional amplitude low-frequency fluctuation, a and DC (degree centrality, b towards the whole brain. In the panel A, this illustrates the entire processes for the calculations of the fALFF: all the brain maps were firstly re-sampled into the 3 mm3 standard space, and corresponding time-series of each voxel would be transformed from time-domain to frequency-domain using the Fast Fourier Transformation (FFT) to obtain the frequency spectrum. Above operations are all done in the corresponding time window, and were showed as the voxel-wise matrices ignoring the space information. In the middle part of this panel, the boxplot shows the dynamics of the fALFF in several sliding-windows to check the temporal fluctuations across participants. To quantify the temporal fluctuation for the dynamic fALFF, a series of fALFF maps were generated for each participant across all the scanning span. Finally, this nature was calculated as the ratio of the average value of fALFF in each voxel on the standard deviation of the corresponding element, which called Coefficient of Variation (CV). For the panel B, this to illustrate the estimation of the dynamic DC in the voxel-wise connectome of whole brain. Firstly, the analysis for the Pearson’s product-moment correlation between each pair of the voxel was conducted in each sliding-time window to generate connectivity matrix. Then, the graph-theoretical analysis was further utilized to calculate the DC for each voxel in light of the threshold at the r > 0.25 or p < 0.05. Finally, the CVs of voxels were estimated to produce the dynamic DC maps

Machine learning

Pattern Recognition for Neuroimaging Toolbox (PRoNTo, v2.0.1; http://www.mlnl.cs.ucl.ac.uk/pronto/prtsoftware.html) was undertaken for pattern classification between HP and LP in this hybrid brain model containing these fusion brain features (Schrouff et al. 2013). Here, both structural entities (GMV images and FA images) and functional schemes (static fALFF maps, dynamic fALFF maps, static DC maps, and dynamic DC maps) were adopted as training features. For feature selection, the whole-brain template derived from LONI Probabilistic Brain Atlas (LPBA40) was used as the first-level mask (http://www.loni.ucla.edu/Atlases). This mask is applied to maximize the probability template and constrain the target area in the whole-brain for selecting useful features. Given the high-dimensional data of these training features, the Gaussian linear kennel algorithm was carried out for shifting from raw pattern matrix (v × n) to well-conditional kernel matrix (n × n) (LaConte et al. 2005; Jiang et al. 2016). No detrending and scaling processes were conducted for all the selected features.

Given the major purposes of current study, the Support Vector Machine (SVM) and L1-Multiple Kernel Learning (L1-MKL) algorithm were applied as classifiers for the recognition of the HP and LP respectively. The SVM was implemented using the toolkit of the LIBSVM (http://www.csie.ntu.edu.tw/~cjlin/libsvm/). In general, the SVM aimed to pursue the hyper-plane (wTx + b = 0) for achieving the linear separation in the high-dimensional space. In this vein, the support vectors of hyper-plane should be found at the maximum 2/|w|once the w*x + b = 1 and w*x + b = -1 (see Fig. 2a) (Månsson et al. 2015; da Rocha et al. 2018). Besides that, the L1-MLK could contribute to performance estimation of multiple predefined kernel for each modality. Owing to the sparsity of this classifier, our findings for each separate feature (modality) could be interpretable in this hybrid brain model (Schrouff et al., 2013). To boost performance of classifier in classification, the hyper-parameters—namely soft-margin maximization (C)—were estimated for optimization in the nested leave-one-subject-out cross-validation (CV) scheme. Specific CV algorithms and how to estimate the performance of classifiers can be found in the SI Method.

Fig. 2.

Fig. 2

Open in a new tab

Diagram of the algorithms of both the support vector machine (a) and mCCA + joint ICA (b). For the sub-graph A, the left panel indicates the estimations for the maximum-margin hyperplane in terms of the support vectors (x); the right panel shows the hybrid brain model of the multi-features, including the brain anatomical properties (Grey Matter Volume [GMV, v1] and Fractional Anisotropy [FA, v2]), static functional properties (fALFF [sfALFF, v3] and Betweenness Centrality [sDC, v4]), and dynamic functional properties (fALFF [dfALFF, v5] and Betweenness Centrality [dDC, v6]). Kernel function is used for the construction of these selected features. For the sub-graph B, the left panel demonstrated the brief framework of the mCCA, and the right panel illustrated the major processes of the joint ICA for the extractions of the joint independent component (Cn). An indicates the matrices of mixing profiles and the corresponding Sn shows matched associated maps. The maximum associations among these selected features would be concatenated as the canonical variables (CV) for elimination and control

mCCA + jICA for fusion data

Owing to the advances of the multivariate analysis technique, the hidden linage information in fusion brain data could be captured effectively by using mCCA + jICA algorithm (Sui et al. 2013a, b). As illustrated in the Fig. 2b, the brief aim of mCCA + jICA was to decompose the associated maps (Ci, i = 1, 2, …, 6) derived from mixing profiles (Canonical Variants, CVi, i = 1, 2, …, 6) of correlation maximization to components (Si, i = 1, 2, …, 6) of remained mixture. In this vein, the hidden or shared information of cross-profiles could be defined as the joint-discriminative independent components (Lohmann et al. 2010; Lottman and White 2017; Lottman et al. 2018). Specific processes and parameters used in this study have been sorted in SI Method.

Results

All raw data, resultant files and interactive NFITI images have been submitted at the OSF repository (https://doi.org/10.17605/OSF.IO/G5S9Q) for open accesses.

Spatial patterns of these features in light of independent modality

To gain understandings of these brain features, one sample t-tests were conducted to generate T-maps for each modality. This process favors the initial understanding of the spatial pattern of these brain modalities for both HP and LP. Results of brain structural features demonstrated the prominent increased grey matter volumes (GMV) in frontal lobe and cingulate cortex for HP (see Fig. 3a). As for neurofunctional features, powerful spontaneous activity and substantial connectomes in temporal and parietal lobes were observed for both HP and LP (see Fig. 3a). Furthermore, the Brainconnectome atlas was used for parcellation of whole brain into intrinsic brain large-scale networks to probe the network-based neural pattern. Results showed that the network-based neuroanatomical pattern of HP were largely comparable to LP, whereas the local and global features in HP were found slightly increased in frontal-parietal system compared to LP (see Fig. 3b).

Fig. 3.

Fig. 3

Open in a new tab

Results for the one-sample t-test pertaining to all the brain features. Derived from one-sample t-test, A shows the spatial maps for all the brain features in both healthy control and procrastinators, including grey matter volume (GMV), fractional anisotropy (FA), static fractional amplitude of low frequency (staticfALFF), static degree centrality (staticDC) and their dynamic unities. B shows the radar plots for all the brain features in intrinsic large-scale brain networks (extending to global brain network). These networks were identified using Brainconnectome atlas, frontal lobe network (FLN), temporal lobe network (TLN), parietal lobe network (PLN), insular lobe network (ILN), limbic lobe network (LLN), occipital lobe network (OLN), subcortical nuclei network (SNN), and global brain network (GBN). All the axis indicate the maximum T value for corresponding network. HP = high trait procrastination; LP low trait procrastination

Revealing neural benchmarks of HP in mass-univariate analysis

To gain further insight into significant results with mass-univariate analysis, independent sample t-tests were carried out to examine the differences between HP and LP for each brain feature (Gaussian Random Field (GRF) correction at voxel-wise of p < 0.001 and cluster-wise of p < 0.05). Given the increasing concerns for the Null-Hypothesis Significant Test (NHST) (Rouder et al. 2009; Masson 2011), aside from above examinations, the Bayesian factor examinations were used for statistical validations as well (see SI Results). These findings showed that the increased GMV of dlPFC, static fALFF of lateral PFC, and DC of dlPFC for HP than do of LP, whereas they revealed the decreased GMV in PHC/OFC and static fALFF in vmPFC/PHC for HP than of LP (see Table S1 and Figure S1-8). These findings were found to be highly parallel to previous studies, showing satisfactory reliability of brain data used in the current study.

Identifying HP from LP accurately by using hybrid brain model

SVM of Leave-One-Subject-Out (LOSO) cross-validation was adopted to evaluate the performance to identify HP and LP in the hybrid brain model. Results showed the balanced accuracy (BAC) of 87.04% (p = 0.003, non-parametric permutation test, n = 1,000) with a sensitivity of 86.42% (true positive for HP, SEN) and specificity of 85.19% (true negative for LP, SPE) for the classification between HP and LP. Area under curve (AUC) of receiver operating characteristic (ROC) curve was determined of 0.85, indicating the high-performance of this classifier. SVM performance tested by tenfold cross-validation supported above findings as well (BAC = 70.06, SEN = 67.08, SPE = 73.05, AUC = 0.74, p = 0.01, non-parametric permutation test, n = 1,000) (see Figure S23).

Furthermore, given the clinical potentials for diagnosis in the future and principle of Occam's razor (Balasubramanian 1997), a simplified hybrid brain model with the fusion brain data of GMV and fALFF was built so as to compromise algorithmic accuracy and efficiency. Typically, the calculations for these brain features would spend a plenty of time. By comparing the duration for processing these features, the GMV and static fALFF were found most efficient. In addition, by using sole brain modality, the GMV, FA and static fALFF were found to outperform classification than other features. Taken them together, the brain features derived from GMV and fALFF were used to built the simplified hybrid brain model. Nevertheless, there was marginally acceptable accuracy of 68.52% (LOSOCV, p = 0.059, non-parametric permutation test, n = 1,000) for identification of HP in this simplified hybrid brain model (sensitivity of 66.67% and specificity of 66.05%) Such findings have been validated by tenfold CV scheme as well (tenfold CV, BAC = 61.01, SEN = 59.15, SPE = 62.87) (see Figure S9-11).

Results derived from L1-MKL model are in support of our hypotheses highly, showing null effects for identifying HP by using each single modality of brain features. In other words, fusion brain information rather than sole brain feature achieved the accurate identifications of trait procrastination (see Table 2). In addition, the voxel-wise discriminative maps for hybrid brain model by using fusion brain data had been provided as well, which indicated the remarkable contributions of frontal lobe and extensive subcortical areas involving into emotional representations (see Figures S12-22).

Table 2.

Performance of classifiers for different features (modalities) in machine learning

Features Accuracy Accuracy bal Sensitivity Specificity p value uncorrected
Hybrid 87.04% 85.80% 86.42% 85.19% .003
Simplified model 79.97 68.52 66.67 68.52 .059
GMV 67.28% 66.05% 48.15% 51.58% .069
FA of white matter 69.14% 68.52% 66.05% 64.81% .052
Static fALFF 53.09% 55.56% 48.77% 46.29% .370
Static DC 50.00% 50.00% 50.00% 43.21% .518
Dynamic fALFF 46.91 49.38% 43.83 45.06 .588
Dynamic DC 47.53 49.38% 43.83 45.06 .562
Open in a new tab

The feature(s) have been highlighted in bold font if it remains p < .05 after FWE correction

Fusion data pattern of hybrid brain model

To reveal the neural pattern of this hybrid brain model in identifying HP, the mCCA + jICA algorithm was performed. Two-sample t-tests were conducted for the hybrid neuroimaging data towards their mixing coefficients between HP and LP. Results of mCCA + jICA revealed four joint-discriminative ICs for the significant differences on loading parameters for HP and LP, including GMV-sfALFF covariations in salience network, FA-sDC-dDC covariations in frontoparietal network, sfALFF-dDC covariations in medial temporal lobe and posterior parietal lobe, and dfALFF-dDC covariations in vmPFC (see Fig. 4 and Table 3).

Fig. 4.

Fig. 4

Open in a new tab

Displays of all the identified joint-discriminative independent components (ICs). There are a total of four joint-discriminative ICs. A and C show the IC3 of GMV and static fALFF respectively; B and D show the IC4 of static fALFF and dynamic DC respectively; Four slices of z =  − 20 to z = 15 are used to present IC9 of FA, static DC, and dynamic DC respectively in this joint component; Six slices of z =  − 20 to z = 20 are adopted to show IC5 of dynamic fALFF and DC respectively in this joint component. Colorbars indicate the z-scores of corresponding IC

Table 3.

P-values of all the features in all of components

Component 1 2 3 4 5 6 7 8 9
GMV 0.20614 0.40967 0.04084 0.09617 0.19672 0.22539 0.42449 0.33638 0.29535
FA of white matter 0.10236 0.44484 0.32499 0.83038 0.90738 0.63456 0.63456 0.39839 0.03137
Static fALFF 0.55477 0.83461 0.01242 0.00705 0.21342 0.22368 0.08469 0.83058 0.15872
Static DC 0.08350 0.72008 0.29281 0.43345 0.31639 0.12052 0.29815 0.86190 0.00851
Dynamic fALFF 0.19419 1.35 × 10 −24 0.22044 0.17047 6.94 × 10 −6 0.29802 0.11335 0.00179 0.60444
Dynamic DC 0.80163 0.20633 0.66930 0.0043 0.00071 0.78061 0.00048 0.12838 0.00142
Open in a new tab

The component(s) have been highlighted in bold font if it (they) is (are) significant for the differences between HP and LP on corresponding brain features

In addition, these modal-specific discriminative ICs of dynamic fALFF (IC2 and IC8, p < 0.001) and dynamic DC (IC7, p < 0.001) were further observed. The modal-specific discriminative IC2 appeared to abnormal dfALFF in PHC and the prefrontal cortex for HP. IC8 indicated predominant aberrations in temporal and parietal lobes in HP. Lastly, IC7 implied the disruptive dDC of vmPFC and OFC for HP compared to LP (see Fig. 5). Full findings can be found in SI Results.

Fig. 5.

Fig. 5

Open in a new tab

Results for all of three modal-specific discrinminative independent components (ICs). Each column is generated with three sub-graph, including brain spatial pattern of IC, line chart (shadow represents the range of loading parameters), and histogram plot for contrast of LP to HP. Colorbars of first row reflect the z-score

Discussion

One major purpose of this study is to identify HP and LP accurately by using neurobiological brain information. To this end, a hybrid brain model was built using the fusion of both structural and functional brain features. Drawing upon machine learning, we estimated and quantified the performance of this model for discriminating HP and LP. Encouragingly, accuracy in identifying HP was obtained by using this hybrid brain model with fusion data—accuracy up to 87%. Additionally, this study also clarified what fusion neural spatial patterns are in this model. Thus, the data-driven mCCA + jICA analysis identifying fusion components of neural pattern was carried out. It showed the crucial roles of shared information in the fronto-temporal cortex, subcortical rewarding circuit, and the PHC in identifying HP. Overall, this study provided a robust hybrid brain model for identification of HP by using fusion data and further unveiled neural spatial components for this model. Thus, these findings shed light on the neuromarkers for identification of trait procrastination.

SVM yielded the high-performance for this hybrid brain model with high accuracy of 87.04% (Sensitivity of 86.42% and Specificity of 85.19%). To obtain further insights into the performance of machine learning for the diagnosis of procrastinators using single brain modality, each brain feature was adopted to generate classifier for determining what the accuracy is for this goal. In this vein, results produced from L1-MKL machine learning demonstrated that the overall accuracy of each model was declined dramatically. The hybrid brain model, which integrated not only brain features structurally and functionally but also brain intrinsic organization locally and globally into one framework, provided a comprehensive understanding of the neural markers that diagnosed procrastinators accurately. It is a striking strength for this study.

Till now, a growing body of literature has been focusing on the accurate identification of disorders by using biomarkers, such as schizophrenia (Koch et al. 2015; Kim et al. 2016; Zarogianni et al. 2017) and Alzheimer's disease (Fu et al. 2019). Nevertheless, existing researches largely relied on a single modality, which was found to be oversimplification for clinical diagnoses (Davatzikos et al. 2008; Woo et al. 2017). To this end, it could be promising and recommended for the combination of multiple modalities diagnosis of neuropsychiatric disorders (Zhang et al. 2012; Peruzzo et al. 2015). Nevertheless, what we understand today about these neurofunctional or neuroanatomical targets for disorders has largely relied on a single modality, and are thus an oversimplification for diagnostic purposes (Davatzikos et al. 2008; Woo et al. 2017). The current study supports the fusion of this hybrid brain model instead of using of any single modality; employing this approach could enable more accurate diagnoses for procrastinators.

To gain further insights into brain spatial pattern of fusion data for the diagnosis of procrastinators, the N-way mCCA + jICA algorithms were drawn. In term of joint-discriminative and modal-specific components, jIC3 and sIC7 shared overlaps in bilateral insula, amygdala, vmPFC, OFC and hippocampus, which were preferentially involved in emotional regulation and episodic prospection towards negative entities substantially (Arce et al. 2008; Shah et al. 2009; Simmons et al. 2011). As aforementioned in the introduction section, potential hypercoupling of emotional regulation and episodic future thinking could be recognized as crucial cognitive pathway for procrastinators (Zhang et al. 2019a, b). In other words, the frequent negative-oriented thoughts might pose individuals to the failure of mood-repair regulation, thus resulting in critically problematic procrastination (Sirois and Pychyl 2013; Eckert et al. 2016; Zhang et al. 2019a, b). Notably, the present study provided the first direct evidence that the multi-modalities neuro-networks supporting emotion regulation and episodic prospection might be the hallmarks of trait procrastination, in accordance with our theoretical framework.

Another contribution worthy to note for the current study is to provide robust evidence supporting the directed brain-behavioral influence of these brain networks to procrastination. Generally speaking, the neural substrates of procrastination were explored by behavioral-brain configuration by mass-univariate analysis. In detailed, researchers frequently regressed procrastination trait to whole-brain voxels, and further did a reverse inference to claim how these brain regions influenced procrastination (e.g., Hu et al. 2018; Liu et al. 2017; Zhang et al. 2016). However, such reverse inferences may suffer from risks of overreaching as it just demonstrates undirected correlational association (Hwang et al. 2020; Lieberman et al. 2019; Kriegeskorte et al. 2009; Vul and Pashler 2012). In this vein, by using hierarchical machine-learning analysis to encode brain information, the current study provided robust evidence to clarify the directional brain-behavioral association of these neural components to procrastination. In other words, this study would facilitate to reveal (not “infer”) that the brain structural and functional co-variances in frontoparietal network (FPN), salience network (SAN), and hippocampus-related network (HRN) could be the neural substrates for producing procrastination. On the flip side, the current study provide robust evidence to substantiate the function-specific brain systems of procrastination. In the proof-of-concept study (Chen et al. 2019), the triple brain network model was proposed by using neuroanatomical features correlating with procrastination theoretically, including self-control, emotional regulation and episodic thinking network. Nevertheless, interpreting these findings into independent brain systems were largely ground theoretical conjectures. Thus, the current study provided substantial evidence to move forward our understanding of procrastination that the procrastination would be produced by multiple neural/psychological pathways independently indeed.

Further, jIC9, sIC2 and sIC8 jointly described the widespread co-variants in the frontal lobe, fronto-parietal network in both brain functional and structural features. Typically, both frontal lobe (especially in dorsal parts) and frontoparietal cortex were well-validated as hallmarks for cognitive control, irrespective of in brain structural or functional configurations (Andrews-Hanna et al. 2014; Dong et al. 2015). Furthermore, more straightforward evidence have revealed that the aberrant downregulation from cognitive control to subcortical dopamine-coded impulsivity may result in out-of-control procrastination (Van Eerde 2000; Ferrari 2001; Rakes and Dunn 2010). Thus, as transdiagnostic biomarkers for numerous psychiatric disorders relating poor cognitive control (e.g., addiction, impulsive-aggressive disorder and etc.), this study allowed us to infer the deficits of cognitive control circuit as one of the mechanism of trait procrastination.

In the current study, there are a few limitations. While the overarching goal was to build a hybrid brain model for the diagnosis of the procrastinators, but the screening of this cohort primarily concentrates on the empirical criterion of epidemiology. In this vein, the clear clinical phenotype for identification of procrastination still lacks a sound available diagnostic criterion. In addition, the trained classifier was not tested by a completed independent sample. Thus, there were no evidence enough to claim that this hybrid brain model could generalize to all the participants. In this vein, expanding the conclusion of this research to widespread domain should warrant some cautions. On the other hand, despite a satisfactory performance for the diagnosis of procrastinators using this sophisticated hybrid brain model, the time cost for the estimations or calculations of these brain features was significantly high. Given the potentials for clinical diagnosis, this work still need to strive to build a practical model for tradeoff between accuracy and availability well (Lv et al. 2016; Fatima and Pasha 2017). In addition, as the intrinsic shortcoming in the machine-learning model, the poor interpretability caused by “blackbox” pattern limited the theoretical contributions in the current study indeed. Thus, it is quite merit to build a interpretable diagnostic model for extending our theoretical understanding of clinical procrastination in the future studies.

In conclusion, this study developed a hybrid brain model for the diagnosis of procrastinators (HP) using fusion neuroimaging data, including brain structural architectures and local/global functional features. To examine and quantify the performance of this hybrid brain model for classifying procrastinators, this potent algorithm—pattern classifications of machine learning was carried out. Further, the data-driven mCCA + jICA scheme allowed us to untangle the brain spatial pattern of fusion data as well. Encouragingly, results for accuracy of machine learning on discrimination between HP and LP can be up to 87.04% at significant level in permutation test. Further, a portion of joint-discriminative independent components were captured from the fusion neuroimaging data, whereby evincing the anomalies of fronto-parietal network, subcortical reward-related encoding, and hippocampus among procrastinators. Collectively, the current study provided a hybrid neural model to identify trait procrastination with high accuracy, and further revealed the neurobiological pattern of trait procrastination.

Supplementary Information

Below is the link to the electronic supplementary material.

Supplementary file1 (DOCX 20549 kb) (20.1MB, docx)

Funding

This work was supported by the National Natural Science Foundation of China (31971026, 31571128), the Fundamental Research Funds for the Central Universities (SWU1809357) and the ChongQing Innovative Research Funds for Graduates (CYB21082).

Data availability

All of raw data, resultant files and interactive NFITI images have been submitted at the OSF repository (https://doi.org/10.17605/OSF.IO/G5S9Q) for open accesses.

Declarations

Conflict of interest

All the authors report no potential conflicts.

Ethical approval

All procedures performed in studies involving human participants were in accordance with the ethical standards of the Institutional Review Board (IRB) of the Southwest University and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Informed consent

Informed consent was obtained from all individual participants included in the study.

Footnotes

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Zhiyi Chen and Rong Zhang have been contributed equally to this work.

References

  1. Andrews-Hanna JR, Smallwood J, Spreng RN. The default network and self-generated thought: component processes, dynamic control, and clinical relevance. Ann N Y Acad Sci. 2014;1316:29. doi: 10.1111/nyas.12360. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Antonucci LA, Penzel N, Pergola G, Kambeitz-Ilankovic L, Dwyer D, Kambeitz J, Haas SS, Passiatore R, Fazio L, Caforio G (2019) Multivariate classification of schizophrenia and its familial risk based on load-dependent attentional control brain functional connectivity. Neuropsychopharmacology 1–9 [DOI] [PMC free article] [PubMed]
  3. Arce E, Simmons AN, Lovero KL, Stein MB, Paulus MP. Escitalopram effects on insula and amygdala BOLD activation during emotional processing. Psychopharmacology. 2008;196:661–672. doi: 10.1007/s00213-007-1004-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Balasubramanian V. Statistical inference, Occam's razor, and statistical mechanics on the space of probability distributions. Neural Comput. 1997;9:349–368. [Google Scholar]
  5. Buckner RL, Sepulcre J, Talukdar T, Krienen FM, Liu H, Hedden T, Andrews-Hanna JR, Sperling RA, Johnson KA. Cortical hubs revealed by intrinsic functional connectivity: mapping, assessment of stability, and relation to Alzheimer's disease. J Neurosci. 2009;29:1860–1873. doi: 10.1523/JNEUROSCI.5062-08.2009. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Casorso J, Kong X, Chi W, Van De Ville D, Yeo BT, Liegeois R. Dynamic mode decomposition of resting-state and task fMRI. Neuroimage. 2019;194:42–54. doi: 10.1016/j.neuroimage.2019.03.019. [DOI] [PubMed] [Google Scholar]
  7. Cercignani M, Inglese M, Pagani E, Comi G, Filippi M. Mean diffusivity and fractional anisotropy histograms of patients with multiple sclerosis. Am J Neuroradiol. 2001;22:952–958. [PMC free article] [PubMed] [Google Scholar]
  8. Chen Z, Liu P, Zhang C, Feng T (2019) Brain morphological dynamics of procrastination: the crucial role of the self-control, emotional, and episodic prospection network. Cerebral Cortex [DOI] [PubMed]
  9. da Rocha JLD, Coutinho G, Bramati I, Moll FT, Sitaram R (2018) Multilevel diffusion tensor imaging classification technique for characterizing neurobehavioral disorders. Brain Imaging Behav 1–12 [DOI] [PubMed]
  10. Davatzikos C, Fan Y, Wu X, Shen D, Resnick SM. Detection of prodromal Alzheimer's disease via pattern classification of magnetic resonance imaging. Neurobiol Aging. 2008;29:514–523. doi: 10.1016/j.neurobiolaging.2006.11.010. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Dong G, Lin X, Potenza MN. Decreased functional connectivity in an executive control network is related to impaired executive function in Internet gaming disorder. Prog Neuropsychopharmacol Biol Psychiatry. 2015;57:76–85. doi: 10.1016/j.pnpbp.2014.10.012. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Eckert M, Ebert DD, Lehr D, Sieland B, Berking M. Overcome procrastination: Enhancing emotion regulation skills reduce procrastination. Learn Individ Differ. 2016;52:10–18. [Google Scholar]
  13. Esteban O, Markiewicz CJ, Blair RW, Moodie CA, Isik AI, Erramuzpe A, Kent JD, Goncalves M, DuPre E, Snyder M. fMRIPrep: a robust preprocessing pipeline for functional MRI. Nat Methods. 2019;16:111. doi: 10.1038/s41592-018-0235-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Fatima M, Pasha M. Survey of machine learning algorithms for disease diagnostic. J Intell Learn Syst Appl. 2017;9:1. [Google Scholar]
  15. Ferrari JR. Psychometric validation of two procrastination inventories for adults: Arousal and avoidance measures. J Psychopathol Behav Assess. 1992;14:97–110. [Google Scholar]
  16. Ferrari JR. Procrastination as self-regulation failure of performance: effects of cognitive load, self-awareness, and time limits on ‘working best under pressure’. Eur J Pers. 2001;15:391–406. [Google Scholar]
  17. First MB, Spitzer RL, Gibbon M, Williams JB. Structured clinical interview for DSM-IV-TR axis I disorders-non-patient edition. New York: New York State Psychiatric Institute; 2001. [Google Scholar]
  18. First M, Spitzer R, Gibbon M, Williams J (1996) Structured clinical interview for DSM-IV axis I disorders, non-patient edition (SCID-NP). Biometrics Research Department, New York State Psychiatric Institute
  19. Friston K, Josephs O, Zarahn E, Holmes A, Rouquette S, Poline J-B. To smooth or not to smooth?: Bias and efficiency in fmri time-series analysis. Neuroimage. 2000;12:196–208. doi: 10.1006/nimg.2000.0609. [DOI] [PubMed] [Google Scholar]
  20. Fu Z, Iraji A., Caprihan A, Adair JC, Calhoun VD (2019) In search of multimodal brain alterations in Alzheimer's and Binswanger's disease. NeuroImage Clin 26:101937 [DOI] [PMC free article] [PubMed]
  21. Hu Y, Liu P, Guo Y, Feng T. The neural substrates of procrastination: a voxel-based morphometry study. Brain Cogn. 2018;121:11–16. doi: 10.1016/j.bandc.2018.01.001. [DOI] [PubMed] [Google Scholar]
  22. Hwang YC, Lee YS, Ryu Y, Lee IS, Chae Y. Statistical inference of acupoint specificity: forward and reverse inference. Integr Med Res. 2020;9(1):17–20. doi: 10.1016/j.imr.2020.01.005. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Itahashi T, Yamada T, Nakamura M, Watanabe H, Yamagata B, Jimbo D, Shioda S, Kuroda M, Toriizuka K, Kato N (2015) Linked alterations in gray and white matter morphology in adults with high-functioning autism spectrum disorder: a multimodal brain imaging study. NeuroImage: Clinical 7:155–169 [DOI] [PMC free article] [PubMed]
  24. Jiang C, Zhang H, Ren Y, Han Z, Chen K-C, Hanzo L. Machine learning paradigms for next-generation wireless networks. IEEE Wirel Commun. 2016;24:98–105. [Google Scholar]
  25. Khullar S, Michael A, Correa N, Adali T, Baum SA, Calhoun VD. Wavelet-based fMRI analysis: 3-D denoising, signal separation, and validation metrics. Neuroimage. 2011;54:2867–2884. doi: 10.1016/j.neuroimage.2010.10.063. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Kim H, Chen C-T, Eclov N, Ronzhin A, Murat P, Ramberg E, Los S, Wyrwicz AM, Li L, Kao C-M. A feasibility study of a PET/MRI insert detector using strip-line and waveform sampling data acquisition. Nucl Instrum Methods Phys Res Sect A. 2015;784:557–564. doi: 10.1016/j.nima.2014.12.080. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Kim J, Calhoun VD, Shim E, Lee J-H. Deep neural network with weight sparsity control and pre-training extracts hierarchical features and enhances classification performance: Evidence from whole-brain resting-state functional connectivity patterns of schizophrenia. Neuroimage. 2016;124:127–146. doi: 10.1016/j.neuroimage.2015.05.018. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Koch SP, Hägele C, Haynes J-D, Heinz A, Schlagenhauf F, Sterzer P (2015) Diagnostic classification of schizophrenia patients on the basis of regional reward-related FMRI signal patterns. PloS One 10:e0119089 [DOI] [PMC free article] [PubMed]
  29. Kriegeskorte N, Simmons WK, Bellgowan PS, Baker CI. Circular analysis in systems neuroscience: the dangers of double dipping. Nat Neurosci. 2009;12(5):535–540. doi: 10.1038/nn.2303. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. LaConte S, Strother S, Cherkassky V, Anderson J, Hu X. Support vector machines for temporal classification of block design fMRI data. Neuroimage. 2005;26:317–329. doi: 10.1016/j.neuroimage.2005.01.048. [DOI] [PubMed] [Google Scholar]
  31. Li J, Seidlitz J, Suckling J, Fan F, Liao W. Cortical structural differences in major depressive disorder correlate with cell type-specific transcriptional signatures. Nat Commun. 2021;12(1):1647. doi: 10.1038/s41467-021-21943-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Liao X, Yuan L, Zhao T, Dai Z, Shu N, Xia M, Yang Y, Evans A, He Y. Spontaneous functional network dynamics and associated structural substrates in the human brain. Front Hum Neurosci. 2015;9:478. doi: 10.3389/fnhum.2015.00478. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Lieberman MD, Straccia MA, Meyer ML, Du M, Tan KM. Social, self, (situational), and affective processes in medial prefrontal cortex (MPFC): causal, multivariate, and reverse inference evidence. Neurosci Biobehav Rev. 2019;99:311–328. doi: 10.1016/j.neubiorev.2018.12.021. [DOI] [PubMed] [Google Scholar]
  34. Liu P, Feng T. The overlapping brain region accounting for the relationship between procrastination and impulsivity: a voxel-based morphometry study. Neuroscience. 2017;360:9–17. doi: 10.1016/j.neuroscience.2017.07.042. [DOI] [PubMed] [Google Scholar]
  35. Lohmann G, Margulies DS, Horstmann A, Pleger B, Lepsien J, Goldhahn D, Schloegl H, Stumvoll M, Villringer A, Turner R (2010) Eigenvector centrality mapping for analyzing connectivity patterns in fMRI data of the human brain. PloS One 5:e10232 [DOI] [PMC free article] [PubMed]
  36. Lottman KK, White DM, Kraguljac NV, Reid MA, Calhoun VD, Catao F, Lahti AC. Four-way multimodal fusion of 7 T imaging data using an m CCA+ j ICA model in first-episode schizophrenia. Hum Brain Mapp. 2018;39:1475–1488. doi: 10.1002/hbm.23906. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Lottman K, White D (2017) SU70. Multimodal fusion OF 7 T imaging data using mCCA+ jICA model in first-episode Schizophrenia. Schizophrenia Bull 43:S186 [DOI] [PMC free article] [PubMed]
  38. Lv F, Wen C, Bao Z, Liu M (Eds) Year Published|. Title|, Conference Name|; Year of Conference Date|; Conference Location| Place Published|:Publisher|. Pages p|.
  39. Månsson KN, Frick A, Boraxbekk C-J, Marquand A, Williams S, Carlbring P, Andersson G, Furmark T(2015) Predicting long-term outcome of Internet-delivered cognitive behavior therapy for social anxiety disorder using fMRI and support vector machine learning. Translat Psy 5:e530 [DOI] [PMC free article] [PubMed]
  40. Masson ME. A tutorial on a practical Bayesian alternative to null-hypothesis significance testing. Behav Res Methods. 2011;43:679–690. doi: 10.3758/s13428-010-0049-5. [DOI] [PubMed] [Google Scholar]
  41. Peruzzo D, Castellani U, Perlini C, Bellani M, Marinelli V, Rambaldelli G, Lasalvia A, Tosato S, De Santi K, Murino V. Classification of first-episode psychosis: a multi-modal multi-feature approach integrating structural and diffusion imaging. J Neural Transm. 2015;122:897–905. doi: 10.1007/s00702-014-1324-x. [DOI] [PubMed] [Google Scholar]
  42. Pruim RH, Mennes M, Buitelaar JK, Beckmann CF. Evaluation of ICA-AROMA and alternative strategies for motion artifact removal in resting state fMRI. Neuroimage. 2015;112:278–287. doi: 10.1016/j.neuroimage.2015.02.063. [DOI] [PubMed] [Google Scholar]
  43. Pruim RH, Mennes M, van Rooij D, Llera A, Buitelaar JK, Beckmann CF. ICA-AROMA: a robust ICA-based strategy for removing motion artifacts from fMRI data. Neuroimage. 2015;112:267–277. doi: 10.1016/j.neuroimage.2015.02.064. [DOI] [PubMed] [Google Scholar]
  44. Rakes GC, Dunn KE (2010) The impact of online graduate students' motivation and self-regulation on academic procrastination. J Interact Online Learn 9
  45. Rebetez MML, Rochat L, Gay P, Van der Linden M. Validation of a French version of the Pure Procrastination Scale (PPS) Compr Psy. 2014;55:1442–1447. doi: 10.1016/j.comppsych.2014.04.024. [DOI] [PubMed] [Google Scholar]
  46. Rebetez MML, Rochat L, Barsics C, Van der Linden M. Procrastination as a self-regulation failure: the role of impulsivity and intrusive thoughts. Psychol Rep. 2018;121:26–41. doi: 10.1177/0033294117720695. [DOI] [PubMed] [Google Scholar]
  47. Rouder JN, Speckman PL, Sun D, Morey RD, Iverson G. Bayesian t tests for accepting and rejecting the null hypothesis. Psychon Bull Rev. 2009;16:225–237. doi: 10.3758/PBR.16.2.225. [DOI] [PubMed] [Google Scholar]
  48. Rozental A, Forsell E, Svensson A, Forsström D, Andersson G, Carlbring P. Psychometric evaluation of the Swedish version of the pure procrastination scale, the irrational procrastination scale, and the susceptibility to temptation scale in a clinical population. BMC Psychol. 2014;2:54. doi: 10.1186/s40359-014-0054-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Schrouff J, Cremers J, Garraux G, Baldassarre L, Mourão-Miranda J, Ceditors P, Mourao-Miranda J. PRoNTo: pattern recognition for neuroimaging toolbox. Neuroinformatics. 2013;11:319–337. doi: 10.1007/s12021-013-9178-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Shah SG, Klumpp H, Angstadt M, Nathan PJ, Phan KL. Amygdala and insula response to emotional images in patients with generalized social anxiety disorder. J Psy Neurosci JPN. 2009;34:296. [PMC free article] [PubMed] [Google Scholar]
  51. Simmons AN, Stein MB, Strigo IA, Arce E, Hitchcock C, Paulus MP. Anxiety positive subjects show altered processing in the anterior insula during anticipation of negative stimuli. Hum Brain Mapp. 2011;32:1836–1846. doi: 10.1002/hbm.21154. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Sirois FM. "i'll look after my health, later": a replication and extension of the procrastination–health model with community-dwelling adults. Personality Individ Differ. 2007;43(1):15–26. [Google Scholar]
  53. Sirois F, Pychyl T. Procrastination and the priority of short-term mood regulation: consequences for future self. Soc Pers Psychol Compass. 2013;7:115–127. [Google Scholar]
  54. Steel P. The nature of procrastination: a meta-analytic and theoretical review of quintessential self-regulatory failure. Psychol Bull. 2007;133:65. doi: 10.1037/0033-2909.133.1.65. [DOI] [PubMed] [Google Scholar]
  55. Steel P. Arousal, avoidant and decisional procrastinators: Do they exist? Personality Individ Differ. 2010;48:926–934. [Google Scholar]
  56. Stöber J, Joormann J. Worry, procrastination, and perfectionism: Differentiating amount of worry, pathological worry, anxiety, and depression. Cogn Ther Res. 2001;25:49–60. [Google Scholar]
  57. Sui J, He H, Pearlson GD, Adali T, Kiehl KA, Yu Q, Clark VP, Castro E, White T, Mueller BA. Three-way (N-way) fusion of brain imaging data based on mCCA+ jICA and its application to discriminating schizophrenia. Neuroimage. 2013;66:119–132. doi: 10.1016/j.neuroimage.2012.10.051. [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. Sui J, He H, Yu Q, Chen J, Rogers J, Pearlson GD, Mayer A, Bustillo J, Canive J, Calhoun VD. Combination of resting state fMRI, DTI, and sMRI data to discriminate schizophrenia by N-way MCCA + jICA. Front Hum Neurosci. 2013;7:235. doi: 10.3389/fnhum.2013.00235. [DOI] [PMC free article] [PubMed] [Google Scholar]
  59. Svartdal F, Steel P. Irrational delay revisited: examining five procrastination scales in a global sample. Front Psychol. 2017;8:1927. doi: 10.3389/fpsyg.2017.01927. [DOI] [PMC free article] [PubMed] [Google Scholar]
  60. Tuckman BW. The development and concurrent validity of the procrastination scale. Educ Psychol Measur. 1991;51:473–480. [Google Scholar]
  61. van den Heuvel MP, Sporns O. Network hubs in the human brain. Trends Cogn Sci. 2013;17:683–696. doi: 10.1016/j.tics.2013.09.012. [DOI] [PubMed] [Google Scholar]
  62. Van Eerde W. Procrastination: self-regulation in initiating aversive goals. Appl Psychol. 2000;49:372–389. [Google Scholar]
  63. Vul E, Pashler H. Voodoo and circularity errors. Neuroimage. 2012;62(2):945–948. doi: 10.1016/j.neuroimage.2012.01.027. [DOI] [PubMed] [Google Scholar]
  64. Walsh JJ, Ugumba-Agwunobi G. Individual differences in statistics anxiety: The roles of perfectionism, procrastination and trait anxiety. Personal Individ Differ. 2002;33:239–251. [Google Scholar]
  65. Wang J, Wang Z, Aguirre GK, Detre JA. To smooth or not to smooth? ROC analysis of perfusion fMRI data. Magn Reson Imaging. 2005;23:75–81. doi: 10.1016/j.mri.2004.11.009. [DOI] [PubMed] [Google Scholar]
  66. Woo C-W, Chang LJ, Lindquist MA, Wager TD. Building better biomarkers: brain models in translational neuroimaging. Nat Neurosci. 2017;20:365. doi: 10.1038/nn.4478. [DOI] [PMC free article] [PubMed] [Google Scholar]
  67. Wu Y, Li L, Yuan B, Tian X. Individual differences in resting-state functional connectivity predict procrastination. Personality Individ Differ. 2016;95:62–67. [Google Scholar]
  68. Yan C-G, Wang X-D, Zuo X-N, Zang Y-F. DPABI: data processing & analysis for (resting-state) brain imaging. Neuroinformatics. 2016;14:339–351. doi: 10.1007/s12021-016-9299-4. [DOI] [PubMed] [Google Scholar]
  69. Yang H, Long XY, Yang Y, Yan H, Zhu CZ, Zhou XP, Zang YF, Gong QY. Amplitude of low frequency fluctuation within visual areas revealed by resting-state functional MRI. Neuroimage. 2007;36(1):144–152. doi: 10.1016/j.neuroimage.2007.01.054. [DOI] [PubMed] [Google Scholar]
  70. Zarogianni E, Storkey AJ, Johnstone EC, Owens DG, Lawrie SM. Improved individualized prediction of schizophrenia in subjects at familial high risk, based on neuroanatomical data, schizotypal and neurocognitive features. Schizophr Res. 2017;181:6–12. doi: 10.1016/j.schres.2016.08.027. [DOI] [PubMed] [Google Scholar]
  71. Zhang D, Shen D, AsDN I. Multi-modal multi-task learning for joint prediction of multiple regression and classification variables in Alzheimer's disease. Neuroimage. 2012;59:895–907. doi: 10.1016/j.neuroimage.2011.09.069. [DOI] [PMC free article] [PubMed] [Google Scholar]
  72. Zhang W, Wang X, Feng T. Identifying the neural substrates of procrastination: a resting-state fMRI study. Sci Rep. 2016;6:33203. doi: 10.1038/srep33203. [DOI] [PMC free article] [PubMed] [Google Scholar]
  73. Zhang S, Becker B, Chen Q, Feng T. Insufficient task-outcome association promotes task procrastination through a decrease of hippocampal–striatal interaction. Hum Brain Mapp. 2019;40:597–607. doi: 10.1002/hbm.24397. [DOI] [PMC free article] [PubMed] [Google Scholar]
  74. Zhang S, Liu P, Feng T (2019) To do it now or later: The cognitive mechanisms and neural substrates underlying procrastination. Wiley Interdiscip Rev Cogn Sci 10:e1492 [DOI] [PubMed]
  75. Zou QH, Zhu CZ, Yang Y, Zuo XN, Long XY, Cao QJ, Wang YF, Zang YF. An improved approach to detection of amplitude of low-frequency fluctuation (ALFF) for resting-state fMRI: fractional ALFF. J Neurosci Methods. 2008;172(1):137–141. doi: 10.1016/j.jneumeth.2008.04.012. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary file1 (DOCX 20549 kb) (20.1MB, docx)

Data Availability Statement

All of raw data, resultant files and interactive NFITI images have been submitted at the OSF repository (https://doi.org/10.17605/OSF.IO/G5S9Q) for open accesses.

Articles from Cognitive Neurodynamics are provided here courtesy of Springer Science+Business Media B.V.

RESOURCES